vibe: baixar versions anteriors

Release notes: https://thewh1teagle.github.io/vibe/changelog/3.2.2
Signed and notarized on macOS and Windows.

Release notes: https://thewh1teagle.github.io/vibe/changelog/3.2.1
Signed and notarized on macOS and Windows.

What's new? 🎉📣
- 🛠️ Transcription is back on older AMD graphics drivers — since the move to the new engine, machines with a 2020-era AMD driver (Vulkan 1.2, common on Ryzen laptops with Radeon graphics) crashed at the first GPU submit with "vkQueueSubmit: Invalid queue". The driver returns an empty queue handle from vkGetDeviceQueue2; Sona 0.6.4 falls back to vkGetDeviceQueue and the GPU works again. Updating the AMD driver is still a good idea, but no longer required
- 🖥️ Older Intel and AMD CPUs are back — Vibe no longer refuses to transcribe on a CPU without AVX2. Sona ships its CPU engine twice, an AVX2 build and an AVX baseline build, and picks the one your processor can run when it starts. Any x86-64 machine from 2011 on now transcribes; the "Your CPU is not supported" message is gone
- 🎮 Out of GPU memory no longer kills the run — when the graphics card runs out of memory partway through a file, Vibe reloads the model on the CPU, retries that file once, and finishes the rest of the queue on the CPU. That model is remembered as too large for your GPU and loads on the CPU from then on. Thanks to @mikemikimike for the report and the first pass (#1471, #1472)
- 🔍 Errors now say what the engine printed — when Sona dies mid-transcription, the error dialog, the logs, and the bug report carry its last lines instead of a bare "failed to read sona event line", so problems like the one above are identifiable from a report
Signed and notarized on macOS and Windows.
Full Changelog: v3.1.7...v3.1.9

What's new? 🎉📣
- 📜 Scroll back through a transcript while it is still being written — following the live text used to be impossible to escape: scrolling up, or clicking a line to play it, snapped you straight back to the tail on the next segment. The hand-off to the reader keyed off a scroll event that was ignored for 700 ms after any scroll Vibe made itself, and every arriving segment refreshed that window, so during a real transcription it never closed. Following now ends on the gestures you actually make — wheel, trackpad, scrollbar, Page Up — and picks up again when you scroll back to the bottom
- 🎥 Google Meet detection asks for the permission it needs — listed in the 3.1.5 notes, but the change landed just after that tag; this is the first build that has it. Reading a browser's window title requires Screen Recording on macOS, and Vibe checked for it without ever asking. It now requests it once and links straight to the settings pane if you decline. Zoom and Teams never needed it
Signed and notarized on macOS and Windows.
Full Changelog: v3.1.5...v3.1.6

What's new? 🎉📣
- ⚙️ Your settings stick again — every setting reverted to its default on restart in 3.1.1. The file on disk was always correct; the app simply never read it back. If you gave up re-setting your preferences, they are safe now
- 🔍 Failures say why — a sidecar that dies now reports its exit code and signal instead of a bare "sona process died", and a crash mid-transcription reads as a crash rather than "error decoding response body". Whisper's own diagnostics reach the log again, after being silently discarded by a no-op callback
- 📦 Broken model downloads are caught — files are verified before they are used, incomplete ones are detected on startup and offered a re-download, and an interrupted download can no longer masquerade as an installed model
- 🧠 Half the memory to load a model — measured 294 MB → 216 MB on a small model, and roughly one model copy less on large ones. Machines that used to run out of memory mid-load have room now
- 🖥️ A CPU without AVX2 is named — instead of dying anonymously, Vibe says the processor is unsupported and why
- 📱 The phone pairing QR scans — one wrong character in the QR generator misplaced the alignment patterns on any code past a certain size, and the pairing code was just past it. No scanner could read it
- 🎙️ Recordings are 9 dB louder — mixing in system audio was quietly halving the level, then halving it again. Every recording made with system audio selected was far quieter than it should have been
- ⌨️ Dictation defaults to a two-key shortcut — Option+Space on macOS, Ctrl+Space elsewhere, instead of a three-key chord
- 🔗 yt-dlp stops nagging — the update prompt appeared on every launch and forgot "Later" each time. It is now a dismissible toast at most once a week, and it offers the update when a download actually fails, which is when it helps
- 🤖 Install the agent skill — Settings → API & Agents writes it to Claude Code and Codex so it is available in every session, with the transcripts folder and the local API baked in
- 🪟 Windows: the VC++ runtime is version-checked — the installer only asked whether any runtime was present, so machines with an old one were skipped and crashed later with no message
- 🌍 The product name and a mistranslated setting fixed across all 22 languages
Signed and notarized on macOS and Windows.
Full Changelog: v3.1.1...v3.1.2

What's new? 🎉📣
- 🎨 Redesigned the whole app around one drop-to-transcribe screen, with a recents sidebar you can resize and search
- ✍️ Rebuilt transcript editing: click anywhere on a line and the caret lands where you clicked, Enter/Tab/arrows move between lines, Escape reverts
- 🎧 The line being spoken is lit as it plays, the view follows along, and a timestamp click plays from there
- 👀 New view options for the transcript — text size, timestamps, speaker labels and text direction
- 🤖 AI summaries now run from the main window: automatically after a transcription when enabled, or on demand from the transcript's AI menu, rendered as markdown and saved with the transcript
- 🔁 Re-transcribe from the sidebar now asks for language, model and options first
- 📂 One picker takes files and folders together on macOS, and folders keep working by drag and drop everywhere
- ⌨️ The global dictation shortcut is set by pressing the keys, with ready-made combos for the ones macOS reserves
- 🔔 Optional "keep running in the tray": closing the window leaves Vibe running for global dictation, off by default (thanks @eggie04)
- ⚙️ Settings moved out of the browser store into an editable app_config.json — agents and people can read and change it, and Vibe picks up the edit immediately
- 🔢 Every number field in settings became a stepper you can hold, type into or reset, with "Never" and "Auto" spelled out
- 🆙 An update waiting is shown on the sidebar toggle instead of hiding in a menu
- 🔤 Fixed uppercase extensions being ignored, so a folder of .MP3 files is no longer empty (thanks @plmelancon)
- 🐛 In-app bug reports now arrive titled with the actual error instead of "App reports bug"
- 🌍 Added Czech (thanks @feretCZ), Bulgarian and German (thanks @nimdassdev); every one of the 22 languages is complete
- ↔️ RTL fixes throughout: the sidebar stays on the window's left, the player timeline drags the right way, and shortcuts read left to right
Full Changelog: v3.0.23...v3.1.0

What's new? 🎉📣
- 🪟 Fixed Vibe and Sona remaining open after exiting on Windows, which could keep GPU memory occupied and prevent relaunching
- 🧠 Added an Advanced setting to automatically unload inactive transcription models and release RAM/VRAM; set it to 0 to keep models loaded
- ⏱️ Protected active uploads, VAD, diarization, and long transcriptions from idle unloading—the timeout begins only after processing finishes
- ⚡ Improved model reuse so Sona skips reloading an already loaded matching model
- 🌍 Added translations for the model inactivity setting in every supported language
- 🔄 Updated Sona to v0.3.5 with Windows parent-process monitoring and reliable child-process cleanup
Full Changelog: v3.0.22...v3.0.23

What's new? 🎉📣

- 🌍 Add Russian language support for i18n configuration
- 🚀 Enhance model download functionality and set Hebrew-AI model for Hebrew
locale
- 🏎️ Replace largest Ivrit model with faster Turbo model (almost as accurate as
Large v2, better at subtitle segmentation)
- 📝 Update model conversion instructions for improved setup and efficiency
- 🧠 Smarter transcription! You can now control how the model chooses what to
say (sampling strategy + beam size)
- 📁 Transcribe Folder recursively

Full Changelog: https://github.com/thewh1teagle/vibe/compare/v3.0.3...v3.0.4

New Contributors

- @vladkor97 made their first contribution in #559
- @CarlosReyesPena made their first contribution in #572
- @jdsmith104 made their first contribution in #583
- @fatslow made their first contribution in #606

Full Changelog: https://github.com/thewh1teagle/vibe/compare/v3.0.3...v3.0.5